Freamon

* Fixed critical bugs in parameter handling:
* Fixed errors when passing unsupported parameters to topic modeling functions
* Fixed parameter extraction for early_stopping_rounds in LightGBM models
* Added comprehensive test coverage for parameter handling
* Implemented parameter filtering to safely handle unsupported parameters
* Added robust fallback mechanisms for error recovery in topic modeling
* Fixed model attribute access in pipeline integration tests
* Improved Pipeline integration with shorthand model types
* Added new test files for comprehensive validation:
* Added test_shorthand_model_types_integration.py for validating model type support
* Added test_early_stopping_parameter.py for early stopping parameter handling
* Added test_topic_modeling_parameters.py for parameter filtering tests

* Enhanced model type support in hyperparameter tuning:
* Added support for shorthand model types like 'lgbm_classifier' and 'lgbm_regressor'
* Improved error messages with list of supported model types
* Unified handling of direct and shorthand model types in tuning pipeline

* Fixed critical bugs in auto_model functionality:
* Fixed parameter handling in LightGBM classifier to prevent "unexpected keyword argument 'early_stopping_rounds'" error
* Fixed handling of 'sampling_ratio' parameter in topic modeling to prevent "unexpected keyword argument" error
* Added robust error handling and recovery for topic modeling in autoflow
* Improved parameter validation and filtering throughout model creation pipeline
* Enhanced error recovery for text processing with better fallback mechanisms
* Fixed index alignment issues in document-topic distributions

* Fixed critical bugs in deduplication module:
* Fixed handling of categorical columns in evaluation to prevent "Categorical does not support reduction 'sum'" error
* Resolved multiprocessing pickling issues in chunked processing by moving nested functions to module level
* Improved robustness for various data types in deduplication evaluation
* Enhanced parallel processing for better stability with large datasets

* Added enhanced deduplication reporting system:
* Implemented comprehensive HTML reports with detailed statistics and visualizations
* Created Markdown reports summarizing deduplication results
* Added Excel export with detailed duplicate pairs and similarity scores
* Implemented PowerPoint report generation for presentations
* Added special Jupyter notebook rendering for interactive exploration
* Implemented interactive visualizations with threshold sensitivity analysis
* Added business impact analysis for optimal threshold selection
* Created cohesive reporting class with multiple export formats
* Added examples demonstrating the enhanced reporting capabilities

* Added advanced LSH algorithms and evaluation capabilities:
* Implemented SimHash for efficient text similarity
* Added BKTree for edit distance similarity search
* Implemented SuperMinHash for more efficient similarity signatures
* Added evaluation with known duplicate flags for measuring accuracy
* Implemented precision, recall, and F1 score calculations
* Added confusion matrix visualization for duplicate detection
* Created threshold sensitivity analysis for finding optimal thresholds
* Implemented advanced LSH method comparison utility
* Added comprehensive examples for all new functionality